Distance-to-fault power outage notification

1 . A system, comprising:

one or more processors, coupled with memory, to:

receive historical time series data corresponding to a characteristic of electricity detected by a plurality of metering devices located on an electricity distribution grid;

receive reported fault locations corresponding to the historical time series data;

generate, using a machine learning model, based on the historical time series data and the reported fault locations, one or more patterns configured to detect locations of faults on the electricity distribution grid based on time series rates of change of a characteristic of electricity subsequent to detection, by a metering device of the plurality of metering devices, of an indication of a fault on the electricity distribution grid; and

store, in memory, the one or more patterns for subsequent retrieval to detect, responsive to a fault, a location of a metering device on the electricity distribution grid relative to the fault.

2 . The system of claim 1 , wherein each of the one or more patterns comprises at least an incline slope and a decline slope.

3 . The system of claim 2 , wherein a declination of the decline slope is greater than an inclination of the incline slope.

4 . The system of claim 1 , wherein the time series rates of change correspond to a derivative in time of a root-mean-square (RMS) value of one of a voltage signal or a current signal measured by the metering devices.

5 . The system of claim 1 , wherein the historical time series data used to train the machine learning model span a predetermined number of waveform cycles subsequent to detection of the indication of the fault.

6 . The system of claim 1 , wherein the machine learning model generates different patterns for different fault scenarios.

7 . The system of claim 6 , wherein the different fault scenarios comprise at least one of a single-line-to-ground fault or a three-phase fault.

8 . The system of claim 1 , wherein the one or more processors further:

generate, using the machine learning model, different patterns based on whether an upstream circuit breaker trips.

9 . The system of claim 1 , wherein the one or more processors further:

normalize values of the time series rates of change to generate likelihood scores of fault location for respective metering devices of the plurality of metering devices.

10 . The system of claim 1 , wherein the one or more processors further:

train the machine learning model using historical time series data labeled with reported fault locations relative to metering devices of the plurality of metering device; and

update the patterns based on confirmed post-event fault locations.

11 . The system of claim 1 , wherein the one or more processors further:

compare, responsive to detection of a fault, a time series rate of change of a characteristic of electricity detected by a metering device with at least one of the stored patterns;

determine, based at least in part on the comparison, a location of the metering device on the electricity distribution grid relative to a location of the fault; and

provide, via a network, for display via a display device, an indication of the determined location of the metering device relative to the fault.

12 . A method, comprising:

receiving, by one or more processors coupled with memory, historical time series data corresponding to a characteristic of electricity detected by a plurality of metering devices located on an electricity distribution grid;

receiving, by the one or more processors, reported fault locations corresponding to the historical time series data;

generating, by the one or more processors, using a machine learning model, based on the historical time series data and the reported fault locations, one or more patterns configured to detect locations of faults on the electricity distribution grid based on time series rates of change of a characteristic of electricity subsequent to detection, by a metering device of the plurality of metering devices, of an indication of a fault on the electricity distribution grid; and

storing, by the one or more processors, in memory, the one or more patterns for subsequent retrieval to detect, responsive to a fault, a location of a metering device on the electricity distribution grid relative to the fault.

13 . The method of claim 12 , wherein each of the one or more patterns comprises at least an incline slope and a decline slope.

14 . The method of claim 13 , wherein a declination of the decline slope is greater than an inclination of the incline slope.

15 . The method of claim 12 , wherein the time series rates of change correspond to a derivative in time of a root-mean-square (RMS) value of one of a voltage signal or a current signal measured by the metering devices.

16 . The method of claim 12 , wherein the historical time series data used to train the machine learning model span a predetermined number of waveform cycles subsequent to detection of the indication of the fault.

17 . The method of claim 12 , wherein the machine learning model generates different patterns for different fault scenarios.

18 . The method of claim 12 , comprising:

generating, by the one or more processors, using the machine learning model, different patterns based on whether an upstream circuit breaker trips.

19 . The method of claim 12 , comprising:

normalizing values of the time series rates of change to generate likelihood scores of fault location for respective metering devices of the plurality of metering devices.

20 . A non-transitory computer-readable medium storing processor-executable instructions that, when executed by one or more processors, cause the one or more processors to:

receive historical time series data corresponding to a characteristic of electricity detected by a plurality of metering devices located on an electricity distribution grid;

receive reported fault locations corresponding to the historical time series data;

generate, using a machine learning model, based on the historical time series data and the reported fault locations, one or more patterns configured to detect locations of faults on the electricity distribution grid based on time series rates of change of a characteristic of electricity subsequent to detection, by a metering device of the plurality of metering devices, of an indication of a fault on the electricity distribution grid; and

store, in memory, the one or more patterns for subsequent retrieval to detect, responsive to a fault, a location of a metering device on the electricity distribution grid relative to the fault.